Source: TEXAS A&M UNIVERSITY submitted to NRP
UNOCCUPIED AERIAL SYSTEM ENABLED PHENOMIC SELECTION TO DEVELOP IMPROVED SOUTHERN MAIZE HYBRIDS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
AFRI COMPETITIVE GRANT
Reporting Frequency
Annual
Accession No.
1025414
Grant No.
2021-67013-33915
Cumulative Award Amt.
$500,000.00
Proposal No.
2020-03523
Multistate No.
(N/A)
Project Start Date
Jan 15, 2021
Project End Date
Jan 14, 2024
Grant Year
2021
Program Code
[A1141]- Plant Health and Production and Plant Products: Plant Breeding for Agricultural Production
Recipient Organization
TEXAS A&M UNIVERSITY
750 AGRONOMY RD STE 2701
COLLEGE STATION,TX 77843-0001
Performing Department
Soil & Crop Sciences
Non Technical Summary
Plant breeding and agronomic technology improvements have improved yields of some crops nine fold in the last 120 years, meaning nine fold less land is need to produce the same amount of food. This has allowed more land to be used for recreation, urbanization or producing more food fiber and fuel sustainably for a growing human population. Still further improvement of crop selection abilities is needed to make plant breeding more resource-efficient and responsive to environmental changes for a growing population. Unoccupied aerial systems (UASs, aka drones) have proven useful as tools to automate routine measurements in plant breeding and increasingly to measure traits previously impossible or infeasible with manual approaches. New statistical prediction approaches using UAS information collected throughout the growing season could allow accurate selections to be made before harvest, speeding the breeding cycle and extending resources. The overall goal of this project is to evaluate if plant selection using many phenotypes measured by UAS can improve the speed and accuracy of decision-making compared with conventional phenotypic selections; leveraging a public maize (corn, Zea mays L.) breeding program as a case study. Specifically in this project we will collect UAS imagery of large breeding populations weekly throughout the growing season and extract dozens of measurements per flight; develop improved statistical models that identify and minimize error using UAS field measurements over time and space; then apply UAS and grain-based phenomic selection models to select the varieties with the most improved yield. Maize is among the most productive crops in the U.S., but comparatively little improvement has occurred across the southern U.S. Improved lines from tropical germplasm will help southern growers, and also enhance adaptation to stresses expected throughout the entire U.S. under a changing climate. This project will leverage, advance and deploy UAS analysis approaches into public tools. The major outputs will include transformative selection and statistical methodologies that will improve plant breeding in all crops, new maize germplasm for current and future farmers, and possibly discovery of fundamental biological indicators that can help better understand the biological adaptation of maize.
Animal Health Component
80%
Research Effort Categories
Basic
20%
Applied
80%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
2031510108110%
2021510108134%
9011510209033%
2027210108123%
Knowledge Area
203 - Plant Biological Efficiency and Abiotic Stresses Affecting Plants; 901 - Program and Project Design, and Statistics; 202 - Plant Genetic Resources;

Subject Of Investigation
1510 - Corn; 7210 - Remote sensing equipment and technology;

Field Of Science
2090 - Statistics, econometrics, and biometrics; 1081 - Breeding;
Goals / Objectives
The overall goal of this project is to evaluate if phenomic selection using Unoccupied Aerial Systems (UAS, aka drones) can improve the speed and accuracy of decision-making compared with conventional phenotypic selections; leveraging a public maize breeding program as a case study.The objectives of this project are to:Collect UAS imagery of new applied maize breeding populations throughout the growing season and extract useful variables from each flight.Develop improved statistical models using spatial, temporal and relatedness features obtained from UAS, which identify and minimize error in prediction.Apply phenomic selection in maize breeding populations based on UAS and grain yield data, to select for yield improvement and compare with conventional selection methods.
Project Methods
Objective 1:Collect UAS imagery of new applied maize breeding populations throughout the growing season and extract useful variables from each flight.Two populations of new breeding lines will be evaluated in standard hybrid yield trials in the field underoptimal andstress conditions. Agronomic and yield data will be collected. Unoccupied aerial systems (UAS also known as drones) will be flown weekly. Images will be processed to extract tabular data on various features and phenotypes. Decisions will be made to advance hybrids based on performance.The success of this objective is determined by usable and good quality field and UAS data for the other objectives, collected and processed in a timely manner. Objective 2:Develop improved statistical models using spatial, temporal and relatedness features obtained from UAS, which identify and minimize error in prediction. Modern and improved data cleaning and analysis steps will be evaluated and applied to tabular data to maximize data usefulness making estimates suitable for phenomic selection, including (i) error and outlier detection, (ii) temporal growth modeling and growth parameter estimation, (iii) field spatial adjustment, (iv) relatedness modeling (kinship, using genomic or phenomic data), and (v), the actual phenomic prediction. The method of analysis will be coded in R and/or Python, and made into open-source functions and packages for use by other investigators. The code will be hosted on GitHub (https://github.com/) and will be included with the publications. The outputs of this objective will be new statistical models that can quickly identify error and detect outliers, model temporal growth, adjust for any spatial effects, and incorporate identity information into the model, as well as clearly annotated R codes that can be used by almost anyone. Success will be defined by cross-validation prediction accuracy, speed, and ease of use by plant breeders and students.Objective 3: Apply phenomic selection in maize breeding populations, based on UAS and grain yield data, to select for yield improvement and compare with conventional selection methods. Using the data collected in Objective 1 then cleaned and analyzed in Objective 2, the best performing hybrids will be selected to be advanced. Two methods will be used to advance germplasm, 1) a conventional approach using yield and whatever other data is available such as breeder rankings. 2) a phenomic selection approach using predictions from season long UAS collected data. Success will be measured by having predictions and decisions from the UAS approach faster than the conventional approach. Overall success is determined by having a holistic and fair comparison of these two approaches to determine what benefits and drawbacks UAS methods provide in plant breeding. While it is believed likely that the new method and newly developed hybrids will provide substantial advantages, to remain unbiased, overall project success cannot depend on such findings.

Progress 01/15/21 to 01/14/24

Outputs
Target Audience:The target audience for the knowledge generated by this project is primarily public and private sector plant breeders and plant scientists at all levels from senior to graduate student. During this reporting period this audience?was reached through presentations and publications. Changes/Problems: Nothing Reported What opportunities for training and professional development has the project provided?Multiple undergraduate,graduate students and postdoctoral scholars have been analyzing the data in various ways, making presentations and submitting publications. Numerous undergraduate students were trained in field breeding methodology. The PI and his team substantially benefitted from the knowledge and application of professional statisticians (Co-PIs and graduate students) on the project. How have the results been disseminated to communities of interest?Results to date have been communicated formally by presentations at regional, national and international meetings and publications. Results to date have been communicated informally through conversations, discussions and interactions where team members where team members were not presenting. What do you plan to do during the next reporting period to accomplish the goals?This is the completion of the project. We plan to continue to analyze collected data, make presentations, as well as prepare and submit publications. All graduate students on the project have graduated. We plan to write additional proposals to further advance the work.

Impacts
What was accomplished under these goals? Overall, a number of landmark studies were published, submitted and or prepared as well as many presentations given to national and international audiences?that included the project. These outputs connect UAS data collection, processing and analysis with breeding program germplasm and goals. UAS data was collected on these populations. The statistical models continued to advance, although the most complex generalizable model would not converge, so we scaled back to simpler models. Phenomic selection was used to compare UAS estimated breeding values to breeders selection choices (based on appearance and yield).

Publications

  • Type: Journal Articles Status: Published Year Published: 2024 Citation: Adak, Alper, Seth C Murray* and Jacob Washburn. 2024. Deciphering Temporal Growth Patterns in Maize: Integrative Modeling of Phenotype Dynamics and Underlying Genomic Variations. New Phytologist (In press) https://doi.org/10.1111/nph.19575
  • Type: Journal Articles Status: Accepted Year Published: 2024 Citation: Alper, Adak, Seth C. Murray*, Jos� I. Varela, Valentina Infante, Jennifer Wilker, Claudia Irene Calder�n, Nithya Subramanian, Natalia de Leon, Jianming Yu, Matthew A. Stull, Marcel Brun, Joshua Hill, Charles D. Johnson, Oscar Riera-Lizarazu, William L. Rooney, and Hongbin Zhang. 2024. Photoperiod Associated Late Flowering Reaction Nor 1 m; Dissecting Loci and Genomic-Enviromic Associated Prediction in Maize. Field Crops Research (Accepted).
  • Type: Journal Articles Status: Accepted Year Published: 2024 Citation: DeSalvio, Aaron J., Alper Adak, Seth C. Murray*, Diego Jarqu�n, Noah D. Winans, Daniel Crozier, and William Rooney. 2024. Near Infrared Reflectance Spectroscopy Phenomic Prediction Can Perform Similarly to Genomic Prediction of Maize Agronomic Traits Across Environments. The Plant Genome (Accepted)


Progress 01/15/22 to 01/14/23

Outputs
Target Audience:The target audience for the knowledge generated by this project is primarily public and private sector plant breeders and plant scientists at all levels from senior to graduate student. During this reporting period this audience was reached through presentations and publications. Changes/Problems:Due to near record environmental stress (heat and drought) in Texas for 2023, like 2022 (drought and heat). The relevance and quality of the data (low yields), was a substantial challenge. Nevertheless, the data will continue to be used to make breeding decisions and develop insights into maize agronomy, biology, genetics and breeding. Additionally, the statistics graduate student needed to extend their work an additional semester, in part due to the complexity of the model. What opportunities for training and professional development has the project provided?Multiple graduate students and postdoctoral scholars have been analyzing the data in various ways, making presentations and submitting publications. Numerous undergraduate students were trained in field breeding methodology. The PI and his team substantially benefitted from the knowledge and application of professional statisticians (Co-PIs and graduate students) on the project. How have the results been disseminated to communities of interest?Results to date have been communicated formally by presentations at regional, national and international meetings and publications. Results to date have been communicated informally through conversations, discussions and interactions where team members where team members were not presenting. What do you plan to do during the next reporting period to accomplish the goals?As little work is left on this project before the final report, the work left to do is to submit additional publications and have students graduate.

Impacts
What was accomplished under these goals? Overall, a number of landmark studies were published, submitted and or prepared as well as many presentations given to national and international audiences that included the project. These outputs connect UAS data collection, processing and analysis with breeding program germplasm and goals. UAS data was collected on these populations. The statistical models continued to advance, although the most complex generalizable model would not converge, so we scaled back to simpler models. Phenomics selection was used to compare UAS estimated breeding values to breeder's selection choices (based on appearance and yield).

Publications

  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Adak, Alper, Seth C. Murray*, and Steven L Anderson. 2023. Temporal phenomic predictions from unoccupied aerial systems can outperform genomic predictions. G3 Genes | Genomes | Genetics. jkac294 https://doi.org/10.1093/g3journal/jkac294
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Adak, Alper, Steven L. Anderson, and Seth C. Murray*. 2023. Pedigree-Management-Flight Interaction for Temporal Phenotype Analysis and Temporal Phenomic Prediction. The Plant Phenome Journal https://doi.org/10.1002/ppj2.20057
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Adak, Alper, Seth C. Murray*, Claudia Irene Calder�n, Valentina Infante, Jennifer Wilker, Jos� I. Varela, Nithya Subramanian, Thomas Isakeit, Jean-Michel An�, Jason Wallace, Natalia de Leon, Matthew A Stull, Marcel Brun, Joshua Hill, and Charles D Johnson. 2023. Genetic mapping and prediction for novel lesion mimic in maize demonstrates quantitative effects from genetic background, environment and epistasis. Theoretical and Applied Genetics 136: 155-162. https://doi.org/10.1007/s00122-023-04394-y
  • Type: Other Status: Submitted Year Published: 2023 Citation: Murray, Seth C., Alper Adak, Steven Anderson, Aaron DeSalvio, Holly Lane, Shakirah Nakasagga 2022. Method of predicting and discovering individuals' fitness and health from phenomic features and use thereof. Provisional Patent submitted October 2023
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Chatterjee, Sumantra, Alper Adak, Scott Wilde, Shakirah Nakasagga, and Seth C Murray* 2023. Cumulative temporal vegetation indices from unoccupied aerial systems allow maize (Zea mays L.) hybrid yield to be estimated across environments with fewer flights. PLOS ONE 18(1): e0277804 https://doi.org/10.1371/journal.pone.0277804
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Adak, A., Murray, S.C.*, Myeongjong, C., Wong, R., Katzfu�, M. 2023. Phenomic data-driven biological prediction of maize through field-based high throughput phenotyping integration with genomic data. Journal of Experimental Botany 74: 53075326, https://doi.org/10.1093/jxb/erad216
  • Type: Journal Articles Status: Published Year Published: 2023 Citation: Andrew W Herr, Alper Adak, Matthew E Carroll, Dinakaran Elango, Soumyashree Kar, Changying Li, Sarah E Jones, Arron H Carter, Seth C Murray, Andrew Paterson, Sindhuja Sankaran, Arti Singh, Asheesh K Singh. 2023. Unoccupied aerial systems imagery for phenotyping in cotton, maize, soybean, and wheat breeding. Crop Science 63: https://doi.org/10.1002/csc2.21028
  • Type: Other Status: Submitted Year Published: 2023 Citation: Murray, Seth C., Alper Adak, and Aaron DeSalvio Method of Objective Measuring Crop Senescence, Grain Filling Period, and Predicting Yield from Remote Sensing Imagery. Provisional Patent submitted October 2023


Progress 01/15/21 to 01/14/22

Outputs
Target Audience:The target audience for the knowledge generated by this project is primarily public and private sector plant breeders and plant scientists at all levelsfrom senior to graduate student. Durring this reporting period this audiance was reached throughpresentations and publications. Changes/Problems:No major changes beyond the delays caused by the winter nursery as described. A graduate student was doing much of the work and is now being hired in this position as a postdoctoral scholar to continue the work. What opportunities for training and professional development has the project provided?Multiple graduate students and postdoctoral scholars have been analyzing the data in various ways, making presentations and submitting publications. Numerous undergraduate students were trained in field breeding methodology. How have the results been disseminated to communities of interest?Results to date have been communicated formally by presentations at regional, national and international meetings and publications. Results to date have been communicated informally through conversations, discussions and interactions where team members where team members were not presenting. What do you plan to do during the next reporting period to accomplish the goals?We will collect another season of hybrid data, make additional hybrid crosses to test, and further develop our unoccupied aerial system (UAS aka drone) pipeline andstatistical model(s). We hoped that this summer we could have had an a priori and de novo head to head comparison between phenomic selection and conventional phenotypic selection methods, however crosses were not successfully made in the winter nursery as hoped.The project was ahead by 6 months but is now behind by a season because of problems with our winter (off-season) nursery.The selected plots were sent to our winter nursery but an equipment malfunction on the only GPS tractor, coupled with supply chain issues in receiving parts for this tractor, one month later, meant we missed the off-seasonplanting window. These crosses are now being made in our summer nursery. This head to head comparison will therefore be delayed by a summer field season.

Impacts
What was accomplished under these goals? Advanced populations of new hybrids were planted and UAS imagery was collected throughout the growing season. Usefulvariables were extracted for each plot from each flight. A number of statistical models were used and tested to integrate multiple years of temporal data. Spatial and error minimization functions are currently being investigated. Phenomic predictions were madein these maize breeding populations based on extracted features fromUAS data and trained using various machine learning approaches on grain yield. This informed crosses to make for the next season. Conventional phenotypic ratings and yield data only were used by the breeder to inform a competing selection choice. These new hybrids will be evaluated in the following field season to compare selection methods.

Publications

  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Adak, Alper, Seth C. Murray*, Clarissa Conrad, Yuanyuan Chen, Nithya Subramanian, Steven Anderson, Scott Wilde. 2021. Validation of Functional Polymorphisms Affecting Maize Plant Height by Unoccupied Aerial Systems (UAS) allows Novel Temporal Phenotypes. G3: jkab075 https://doi.org/10.1093/g3journal/jkab075
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Lane, Holly, Seth C. Murray*. 2021. High throughput can produce better decisions than high accuracy when phenotyping plant populations. Crop Science https://doi.org/10.1002/csc2.20514
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Adak, Alper, Seth C. Murray*, Steven L Anderson II, Sorin C. Popescu, Lonesome Malambo, M. Cinta Romay, Natalia de Leon. 2021. Unoccupied aerial systems discovered overlooked loci capturing the variation of entire growing period in maize. The Plant Genome, e20102 https://doi.org/10.1002/tpg2.20102
  • Type: Journal Articles Status: Under Review Year Published: 2021 Citation: Aaron J DeSalvio, Alper Adak, Seth C Murray*, Scott C Wilde, Thomas Isakeit. Phenomic Data-Facilitated Rust and Senescence Prediction in Maize Using Machine Learning Algorithms. Research Square, 23 Nov 2021. DOI: 10.21203/rs.3.rs-1108535/v1
  • Type: Journal Articles Status: Under Review Year Published: 2021 Citation: Adak, Alper, Seth C. Murray*, Steven L Anderson. Temporal phenomic predictions from unoccupied aerial systems can outperform genomic predictions. bioRxiv October 08, 2021 https://doi.org/10.1101/2021.10.06.463310
  • Type: Journal Articles Status: Published Year Published: 2021 Citation: Adak, Alper, Seth C. Murray, Sofija Bo~inovi?, Regan Lindsey, Shakirah Nakasagga, Sumantra Chatterjee, Steven L. Anderson, and Scott Wilde. 2021. Temporal Vegetation Indices and Plant Height from Remotely Sensed Imagery Can Predict Grain Yield and Flowering Time Breeding Value in Maize via Machine Learning Regression. Remote Sensing, 13(11), 2141. https://doi.org/10.3390/rs13112141
  • Type: Journal Articles Status: Awaiting Publication Year Published: 2021 Citation: Zhang*, Zhiwu, Chunpeng Chen, Jessica Rutkoski, James Schnable, Seth Murray, Lizhi Wang, Xiuliang Jin, Benjamin Stich, Jose Crossa, Ben Hayes. 2021. Harnessing Agronomics Through Genomics and Phenomics in Plant Breeding: A Review. Plant Breeding Reviews. (in press) doi: 10.20944/preprints202103.0519